A somewhat relevant question: what is a good way to get into longevity research/research support for a software engineer with a degree in optical engineering, who recently started refreshing college math?
An uncle of mine got his PhD and then after his postdoc spent the rest of his working life helping other researchers at his university build equipment to support experiments. He had skills with machining and design that most of his colleagues lacked. Much of hands-on scientific research can be improved by some not-too-complicated piece of equipment, but that equipment isn't available off the shelf yet. Some researchers are lucky enough to be supported by people like my uncle, or already have skills like my uncle. Most biological researchers aren't also engineers. You might advance the productivity of biological research, maybe even do well financially if you design something that a company like Millipore ends up acquiring.
Tedious pipetting work used to be a major waste of time in some kinds of biological research. There are commercial robots for that now. What's the next most tedious thing that could be improved with good tooling? I don't know, but you might want to see if you can find out.
EDIT: I'm suggesting that you look for opportunities regarding experimental research instead of pure software because I'm not sure current experimental data is good/abundant enough. I was peripherally involved with an academic "proteomics" software effort more than a decade ago (is that still a trendy thing?) and my experiences led me to believe that experimental reproducibility and throughput needed to improve before it was worth focusing on software. I also hear biologists gripe about slow, poorly reproducible cell experiments in places like the comments on Derek Lowe's blog.
My personal hobby is computational chemistry but if I wanted to make a real impact on chemistry I think it would have to relate to instrumentation or tooling for bench chemists. Chemistry and especially biology are too complicated for theoretical/computational approaches to contribute much without collaborating with experimentalists.
More specifically, the difficulty is in developing computationally efficient models (i.e. algorithms that could be used on today's computers) - vs. just using computational methods of quantum mechanics, which in theory should be able to model anything that consists of atoms but in practice turns out to be too computationally intensive.
To elaborate on your comments, most of computational chemistry does use quantum mechanical models, and there are indeed difficult problems with computational intensity. Basic quantum chemical methods start with a big-O time complexity of O(N^4). The "gold standard" of computational chemistry, CCSD(T), is O(N^7). It is the worst-scaling method that still sees routine use.
https://en.wikipedia.org/wiki/Ab_initio_quantum_chemistry_me...
An "exact" [1] approach to electronic structure calculations, full configuration interaction, scales as O(N!) -- yes, factorial. Not surprisingly, the size of systems tractable via FCI has not grown much in 30 years even as computers have grown much faster.
There is indeed a lot of work applied developing efficient approximations to the "exact" quantum mechanical solution, and to eking out more constant-factor improvements from existing algorithms.
There's also a lot of work on taking electronic structures, available from various methods, and deriving familiar chemical properties from them. Things like NMR spectra, Raman spectra, pKa, melting point, aqueous solubility...
Measuring properties of bulk condensed-phase matter in the lab is easy but it's hard in simulation. Something "basic" like melting point is very hard to derive from ab initio calculations. On the other hand, properties that require expensive equipment to measure, like NMR spectra, are comparatively easy to calculate.
[1] Terms and conditions apply. Consult Helgaker et al. "Molecular Electronic‐Structure Theory" for details.
[1] https://www.springdisc.com/#careers
[2] https://medium.com/spring-discovery/with-18-million-in-new-f...
Cheers, Peter
We can't really do much about ageing until we solve cancer as ageing is the evolutionary original anti-cancer system.
There are some mutations in people that slow ageing at the expense of increasing the cancer rate. There is a very interesting one from Brazil where a mutation in the p53 gene has this exact mechanism [0].
0. https://youtu.be/URKJ7LLXc3E (you can skip the first 5 minutes - the relevant part is around minute 15).
Heart disease research investments would likely have equal or superior payoffs to cancer research investments.
Yes the life expectancy increase from curing cancer is only around 2 years, but it is the essential first step to doing something major about ageing.
anyway, I don't think anybody in the serious scientific community believes that a cancer-only research program would have a huge impact on longevity and instead, most people advocate for a portfolio with roughly 70% spent across cardiovascular and cancer, and the rest on other causes.
BTW what you've said is also fairly philosophical. It would be completely correct to say that heart disease is a cause of aging, under a reasonable definition of aging.
you shouldn't expect, with your academic pedigree and work experience, to be able to pick up enough biology to be truly useful for deep discovery. You can help out writing code, but don't expect to be able to design, run, and analyze the results of an experiment. In biology, it takes decades to be able to judge the results (very different from computer science and machine learning).
Areas where it won't work: any time you have new image data that doesn't resemble what the networks were trained on. In fact, most people in the field recommend training on and running inference on a single microscope and if you change scopes, you have to retrain your model! Obviously data augmentation has a lot to contribute there but there a ton of challenges.
I've actually proposed building a warehouse-scale microscopy facility within a couple miles of amazon or google data center with full realtime reinforcement learning loop. If you have hundreds of near-identical scopes collecting the same data, you can train over the variation.
The Longevity Investor Network [1] might be of interest. (Disclaimer: I have donated, and actively consider investing, through them.)
(Disclaimer: I'm a regular donor to the former and an investor in the latter, in both cases because I believe in their mission.)
I would look into optical imaging and biophotonics there are many exiting emerging technologies that may revolutionize research (into longevity among others) and healthcare in general. Optical coherence tomography for example.